And-Parallelism Implementation

• By translation to or-parallelism: [Carlsson]
  • Simplicity
  • Relatively high overhead higher need for granularity control
  • Used in ECLPS system.

• Direct implementation:
  • Abstract machine needs to be modified: e.g., PWAM / Marker model or AKL machine.
    • System comprises a collection of agents (processes/processors).
    • Each agent is an LP/CLP engine with a full set of stacks.
    • Scheduling is normally done lazily through goal stacks.
  • Low overhead (but granularity control still useful)
  • Direct support for concurrent/parallel language
  • Used in &-Prolog/Ciao-Prolog system and AKL.

• Issues in direct implementation:
  • Scheduling / fast task startup
  • Memory management
  • Use of analysis information to improve indexing
  • Local environment support (AKL)
  • Recomputation vs. copying
  • Efficient implementation of parallel backtracking
  • Efficient implementation of ask
  • etc.